The neural circuit implementation of distributional reinforcement learning

Distributional reward prediction error signals are preferentially encoded in lateral ventral striatum-projecting dopamine neurons, while reward distribution information is shared between ventral and dorsal striatum.

• Why it matters: It remains unclear how distributional RL’s uncertainty-aware computations are implemented in basal ganglia circuits, limiting understanding of risk-sensitive learning and decision-making.
• What they did: The study combined striatum-wide fiber photometry of dopamine axonal activity with Neuropixels recordings from projection-defined dopamine neurons to map outcome-distribution representations.
• The result: An anatomical gradient of optimism in the midbrain supports diverse value predictions, extending basal ganglia actor-critic ideas by revealing distributional prediction errors in a mesolimbic dopamine subpopulation.

AI summary of the abstract · may contain errors · not medical advice · read the paper.